Estimation of Montreal Cognitive Assessment Scores Using Caregiver Reports and Demographics: A Model Development Study

Background/Objectives: Assessment of cognitive function in patients with dementia is often hindered by functional and environmental barriers. Although caregiver reports are an alternative source, their clinical utility for estimating patients’ cognitive function remains uncertain. This study aimed to estimate cognitive function using caregiver-reported data combined with patient demographics and to evaluate its clinical utility. Methods: This retrospective cross-sectional study enrolled participants who visited a memory clinic and completed the Montreal Cognitive Assessment (MoCA) for cognitive assessment, together with caregiver-reported questionnaires for activities of daily living (ADL) and neuropsychiatric symptoms (NPS). Multivariable linear regression models were constructed to predict the MoCA score, with Model 1 including demographics, ADL, and NPS as covariates and Model 2 further incorporating clinical diagnosis. The intraclass correlation coefficient, Bland–Altman analysis, and regression error characteristic curves were assessed. Results: Among 2650 participants (56.5% women; mean age, 70.4 years), the NPS variable was excluded from both models. Model 1, which included demographics and ADL, explained 65.4% of the variance, whereas Model 2, which incorporated clinical diagnosis, explained 75.9%. Model 2 yielded an intraclass correlation coefficient of 0.853, compared to 0.778 for Model 1. At a 4-point error tolerance, Model 2 yielded an accuracy of 75.5%. Bland–Altman biases were near zero, with 95% limits of agreement of approximately ±7 points for Model 2. Conclusions: MoCA scores can be estimated using caregiver-reported ADL scores, demographics, and clinical diagnosis. NPS scores provided no additional predictive value when these factors were included. These models provide valid quantitative tools for indirect cognitive assessment when in-person testing is impossible.